Fig 1.
Stimulus design of the 12-target BCI system.
(A) The user interface of a virtual keypad for a phone-dialing program. (B) Frequency and phase values specified for each target. The red square in (A) is the visual cue indicating a target symbol ‘5’ in the experiment.
Fig 2.
Performance comparison of seven CCA-based SSVEP detection methods.
(A) Averaged classification accuracy and (B) simulated ITRs across subject using different data lengths. Error bars indicate standard errors. The asterisks indicate significant difference between different methods (*p<0.05).
Table 1.
Classification accuracy (%) for each subject with 1s data length.
Fig 3.
Comparison of accuracy with different parameters.
Averaged classification accuracy for each method using (A) different numbers of harmonics (Nh in Eq (5)) and (B) different number of channels (Nc) across subjects. Error bars indicate standard errors.
Fig 4.
The waveforms of the training data of SSVEPs and sinusoidal reference signals at 12.25Hz after spatial filtering based on CCA for each subject.
Fig 5.
Examples of feature values and r-square values for each method.
(A) Averaged feature values and (B) r-square values for SSVEPs at 12.25Hz. The dotted line in (A) indicates the stimulus frequency. Error bars in each subfigure indicate standard errors. The asterisks indicate significant difference between two different methods.
Fig 6.
Averaged classification accuracy across subjects with different numbers of training trials for each method.
Error bars indicate standard errors.
Table 2.
Averaged computational time (ms) ± standard deviation for single-trial analysis across subjects.
Table 3.
Summary of CCA-based SSVEP Detection Methods.